[Bug math/20916] New: pow handling of sNaN arguments

[jsm28 at gcc dot gnu.org]

https://sourceware.org/bugzilla/show_bug.cgi?id=20916

            Bug ID: 20916
           Summary: pow handling of sNaN arguments
           Product: glibc
           Version: 2.24
            Status: NEW
          Severity: normal
          Priority: P2
         Component: math
          Assignee: unassigned at sourceware dot org
          Reporter: jsm28 at gcc dot gnu.org
  Target Milestone: ---

Various pow function implementations mishandle sNaN arguments in various ways. 
This includes returning sNaN instead of qNaN for sNaN arguments.  For arguments
(1, sNaN) and (sNaN, 0), TS 18661-1 semantics are also that the result should
be qNaN, whereas with a qNaN argument there the result should be 1.

Some representative examples for double for x86_64 follow; the detailed
failures vary between implementations.  This bug specifically excludes any
issues for float and double for 32-bit x86, given the problems with testing
sNaNs in those cases.

Failure: Test: pow (sNaN, 0)
Result:
 is:          1.0000000000000000e+00   0x1.0000000000000p+0
 should be:  qNaN

Failure: Test: pow (sNaN, qNaN)
Result:
 is:         sNaN
 should be:  qNaN
Failure: Test: pow (0, sNaN)
Result:
 is:         sNaN
 should be:  qNaN

Failure: Test: pow (1, sNaN)
Result:
 is:          1.0000000000000000e+00   0x1.0000000000000p+0
 should be:  qNaN

-- 
You are receiving this mail because:
You are on the CC list for the bug.